Why AI Adoption Is Slowing — and Why Control Is The Missing Feature

We’re hearing the same threat on repeat: the AI bubble will burst.

For months, experts have been predicting not if, but how, the carnage will unfold — and now we’re seeing the first blip in AI growth since 2023. Investors are still pouring record capital into AI, but capital doesn’t guarantee adoption. User adoption, not funding, determines survival — and AI adoption rates for large firms are trending down.

Even where adoption remains steady, we’re seeing warning flares. Smaller companies and startups are still consistently using AI, but they’re suffering from something else: use of AI is being linked to burnout.

When it comes to AI adoption by knowledge workers, something is off. AI products are failing us — and the missing link is greater control and customization.

Adoption Apocalypse Now

AI tools should remove busywork, accelerate processes, and make people’s lives better. But there’s a ton of ways that AI tools fall wide of the mark — whether it’s implementation or a mismatch between workflow and goal.

Let’s break down why AI tools are losing their luster for knowledge workers.

Executive Aspirations vs. Employee Reality

From the 10,000ft view, AI looks transformative. Down in the trenches, it’s a different story.

Executives are thrilled by the promised productivity gains of implementing AI, and companies increasingly mandate the use of AI. But individual contributors are the ones who endure the frictions, from new workflows and training, to potential mismatches between executive expectations and IC reality.

Automation vs. Regulation

AI can draft in seconds, but companies can take weeks to deliberate. So it’s no mystery why adoption is faltering at larger companies.

AI is faster than the speed of thought, but a big company’s processes aren’t. They’re built for risk mitigation and legal compliance, and AI can be a liability. Consequently employees have to create new, undesirable workflows, like new review stages to catch AI hallucinations, or picking through the flaws in AI-generated code.

Boundless AI vs. Employee Burnout

There’s another disturbing side-effect: AI burnout. According to a recent study published in the Harvard Business Review, using AI contributes to task expansion, distracted multitasking, and poorer work-life boundaries. 

People feel the need to achieve more, and more, with AI — and that’s like a bullet-train to burnout. Some of the pressure is self-imposed, as we all collectively raise expectations for what we can achieve. But executives can also use the AI revolution as leverage and blame AI for higher performance expectations — and even layoffs. It’s hardly surprising, then, that employees report feelings of disconnection and loss of purpose when AI technology is added to their workflows. 

Misaligned AI

Friction, broken workflows, employees who are exhausted and demoralized… These symptoms all have the same root cause: AI that’s over-inserted and under-customized.

So what if an LLM can produce an impressive-at-first glance report, or churn out surface-level-convincing code? If it generates an extra few hours of work reviewing and fact-checking, it’s failing on its most basic promise of streamlining and speeding up our work. No wonder the novelty is gone, and we’re now seeing an increase in AI hesitancy.

A 2025 study of people in 47 countries reveals that only 46% of people trust the technical ability of their AI tools, while a separate 2025 report shows that 82% aren’t finding effective ways to use AI in their day to day. If you’re putting in work to finagle AI into your workflow, or patching your processes with extra review steps, then you’re working for AI — not the other way around.

Getting Back in the AI Driver’s Seat

What we’re missing is greater choice and better customization.

It seems simple, right? Choosing when and how to use AI turns it from a potential wrecking ball in your company workflows, into a missile that you’re deploying right on target. And tools that allow for better customization empower people to fit AI into their workflows more smoothly.

Yet user controls haven’t always been front-of-mind for Big Tech companies. As many commentators are noting, somewhere along the way AI stopped being optional and became woven into products as default, rather than opt-in. Could that be the moment when users started disengaging, and adoption began slowing down?

The Customization Race

People will continue to reject one-size-fits-all tools and hastily implemented AI workflows. And now we’re seeing AI leaders start to respond.

Some of the big players are improving customization, like ChatGPT’s introduction of personalization features that allow us to play with the tone, energy level, and even emoji volume of our AI chatbots. Some commentators think we’ll see AI tools niche down and win adoption by becoming more and more granular, supporting hyper-specific tasks.

AI That Adapts To You

It’s becoming clear that in the next phase of AI development, the most adaptable tools will win. Tools that do things like this:

• Adapt to existing workflows, instead of forcing you to create new ones. Zoom AI Companion 3.0 has gotten this memo and launched upgrades including a platform-agnostic notes tool — meeting users in the tools they already use. 
• Let employees choose when AI participates, and when it steps back. Like GitLab Duo, where users choose which groups and projects can use AI capabilities.
• Reduce cognitive load, instead of adding a new workflow for checks. ClickUp Brain, for instance, operates inside workflows, not alongside them, embedding AI into task views — like thread summaries, updates, and suggested next steps.
• Provide transparency into how outputs are generated, instead of placing the burden on employees to figure it out. Tools like WorkflowGen focus on AI responsibility by building audit trails directly into AI actions, so human reviewers don’t have to reverse-engineer outputs. 

The New AI Gold Standard

The first wave of AI was driven by possibility. There were dazzling demos, record-breaking investments, and sky-high expectations…but now the novelty has worn off, and the market is recalibrating.

The next wave will be driven by something much more powerful: usability. We’ve seen what happens when AI outpaces integration: burnout, workflow frictions, and, inevitably, slower adoption. These aren’t signs that the AI bubble is bursting, but that AI without sufficient customization is more of a burden than a boost.

The next competitive moat for AI builders will be trust — and trust comes from letting users lead.